zfs discuss - Jun 2010 - does sharing an SSD as slog and l2arc reduces its life span?

Hi,

I don''t know if it''s already been discussed here, but while
thinking about using the OCZ Vertex 2 Pro SSD (which according
to spec page has supercaps built in) as a shared slog and L2ARC
device it stroke me that this might not be a such a good idea.
Because this SSD is MLC based, write cycles are an issue here,
though I can''t find any number in their spec.
Why do I think it might be a bad idea: L2ARC is quite static
in comparison with ZIL and L2ARC takes all the place it can get.
But if 90% of the device are nearly statically allocated, the
devices possibilities for wear-leveling are very restricted.
If the ZIL is heavily used, the same 10% of the device get
written over and over again, reducing the life span by 90%.

Is there some fundamental flaw in this line of thought?

Thanks,
Arne

On Sat, 19 Jun 2010, Arne Jansen wrote:
> If the ZIL is heavily used, the same 10% of the device get
> written over and over again, reducing the life span by 90%.
The SSD design might implement wear leveling which moves an old block 
which was only written once or twice to the block with heavy wear. 
The offsets to the two blocks would then be logically swapped. This 
seems like a reasonable design choice since some blocks are virtually 
read-only.  The SSD is not necessarily restricted to only writing 
blocks on request of the user.
> Is there some fundamental flaw in this line of thought?
No.  Normally you would not want to use a MLC SSD as a slog.  The SLC 
SSDs wear out quicker than one would like under heavy repeated writes. 
Over-provisioning the slog SSD storage size should help.

Bob
-- 
Bob Friesenhahn
bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/

Well pretty much by definition any writes shorten the drives life, the more
writes the shorter it is.

That said, here is some interesting math that I did before I built my first mlc
array.

For a certain brand of indellix drive I calculated the life span in the
following way. Based on the maximum sustained write speed of the drive and the
size of the drive (256GB by the way) it would take 9 months to over write the
entire drive 10000 times at 100% busy writing.

However I knew that my controller would be lucky to keep all the drives at 25%
busy (btw turns out that it''s really about 12%) so I took the 9 months
and multiplied that times 4 coming up with 36 months.

Great now we''re at 3 years, but we''re still doing 100% writes
and we know that this isn''t going to be the case. In fact we expect the
absolute worst case scenario is that we''ll be doing less than 25%
writes. So again I took the 3 years and multiplied that times 4.

This comes out to 12 years to wear out my mlc drives. Just in case I''m
calling it 10 years. But you know what? Quite frankly those boxes will be
retired in less than 5 years and even then I''ll be suprised if
it''s still my problem to worry about.

Of all the issues that might concern me about using mlc drives, them wearing out
isn''t really one of them.

Of course if your useing tiny drives, the math changes. In fact under the above
scenario, assuming a 32gb drive went as fast as a 256gb drive (and they
don''t btw) your 32gb drive would only last about 18 months. Since
it''s probably only has half the chip count of the larger drive, and
thus only using half it''s write channels, you probably still have about
3 years of life in the drive running at 25% busy x 25% writes.

Just some food for thought.
-- 
This message posted from opensolaris.org

On Sat, 19 Jun 2010, Richard Jahnel wrote:
> For a certain brand of indellix drive I calculated the life span in 
> the following way. Based on the maximum sustained write speed of the 
> drive and the size of the drive (256GB by the way) it would take 9 
> months to over write the entire drive 10000 times at 100% busy 
> writing.
Did you consider the ''write amplification'' factor?  For
example, if
the SSD uses a 4K erasure block but only one 512-byte sector is 
updated per write?  Tiny writes can wear out the SSD much faster than 
bulk write rates would suggest.
> Just some food for thought.
More food!

The MLC drives seem to usually have more write latency than the SLC 
drives.

Bob
-- 
Bob Friesenhahn
bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/

TBH write amp was not considered, but since I''ve never heard of a write
amp over 1.5, for my purposes on the 256gb drives they still last welll over the
required 5 year life span.

Again it does hurt a lot when your using smaller drives that less space
available for wear leveling.

I suppose for cache drives it will only be a minor annoyance when you have to
replace the drive. Seeing as a cache failure won''t lead to data loss.
In my mind it would be more of a concern for a slog drive.
-- 
This message posted from opensolaris.org

On 6/19/10 3:56 AM, Arne Jansen wrote:
> while
> thinking about using the OCZ Vertex 2 Pro SSD (which according
> to spec page has supercaps built in) as a shared slog and L2ARC
> device
IMO it might be better to use the smallest (50GB, maybe overprovisioned 
down to ~20GB) Vertex 2 Pro as slog and a much cheaper SSD (X25-M) as L2ARC.
> But if 90% of the device are nearly statically allocated, the
> devices possibilities for wear-leveling are very restricted.
> If the ZIL is heavily used, the same 10% of the device get
> written over and over again, reducing the life span by 90%.
As Bob Friesenhahn said, you''re assuming dynamic wear leveling but 
modern SSDs also use static wear leveling, so this problem doesn''t 
exist. (Note that in this context the terms "dynamic" and
"static" may
not mean what you think they mean.)

Wes Felter

Wes Felter wrote:
> On 6/19/10 3:56 AM, Arne Jansen wrote:
>> while
>> thinking about using the OCZ Vertex 2 Pro SSD (which according
>> to spec page has supercaps built in) as a shared slog and L2ARC
>> device
> 
> IMO it might be better to use the smallest (50GB, maybe overprovisioned 
> down to ~20GB) Vertex 2 Pro as slog and a much cheaper SSD (X25-M) as 
> L2ARC.
> 
No budget for this. Lucky if I can get the budget for the Vertex 2 Pro.
But if this sharing works (thanks to static wear leveling) it should be
sufficient to leave 10-20% space.

> As Bob Friesenhahn said, you''re assuming dynamic wear leveling but
> modern SSDs also use static wear leveling, so this problem doesn''t
> exist. (Note that in this context the terms "dynamic" and
"static" may
> not mean what you think they mean.)
Thanks for the term. Yes, this makes sense.

--
Arne